A new report dentifies enterprise VR demand as a relative bright spot in a shrinking market. Busy fleets and robot-training projects could support more software and services spending without producing a comparable rise in headset orders.
Global VR headset shipments fell 18% year on year in the second quarter of 2026, while enterprise VR demand remained resilient, according to Counterpoint Research’s latest quarterly update.
Shipments also dropped 16% from the previous quarter. Counterpoint attributes some of the pressure to Meta’s headset price increases amid rising memory costs. It describes businesses and governments as less price-sensitive when buying VR to solve operational problems, and expects robotics to generate more demand for data collection and remote control this year, particularly in China.
The public summary doesn’t quantify enterprise shipments or revenue. Its robotics assessment is a forecast, so neither finding establishes that business customers can replace the consumer sales being lost.
Enterprise customers can also expand their use of VR without ordering many more devices. A training team might want its existing fleet working harder, whereas a robot developer wants demonstrations that help machines operate independently. Both can get more value from a headset while giving its manufacturer little additional hardware revenue.
Busy Fleets Don’t Always Mean New Orders
Consider an illustrative training centre with 20 headsets. If better scheduling allows it to increase completed sessions from 200 to 400 a month using the same equipment, participation doubles while its headset purchases remain unchanged.
The organisation might buy additional training modules or pay for integration with its learning-management system. A supplier could earn more from the customer through those services. A subscription charged solely by headset, however, wouldn’t automatically grow with the number of people trained.
For a shared fleet, equipment requirements depend on how many people need to train simultaneously, alongside session length and the time needed to prepare devices between users. Total employee numbers alone are a poor basis for forecasting headset orders.
Counterpoint’s figures don’t establish that higher utilisation caused the shipment decline. They also can’t reveal whether an existing fleet is collecting dust or handling more work than last year.
Suppliers planning around enterprise demand therefore need customer records covering active use, renewals and expansion into further sites. Device orders show where equipment is going. Repeat spending shows whether the customer has found a reason to keep the programme running.
More Robots Don’t Require a Headset Each
Robotics creates another way for VR equipment to serve a much larger operation than its unit count suggests.
As The XR Beat reported earlier this month, Noble Machines uses VR teleoperation to capture human demonstrations for its Moby robots. NVIDIA’s case study describes a development process combining those demonstrations with thousands of simulated robot instances running in parallel.
Noble’s own account describes augmenting demonstration data and using human interventions to correct failures. The company wants successful robot behaviour to generate further training data, reducing reliance on people demonstrating each new example.
The commercial inference is that a growing robot fleet needn’t produce an equally large fleet of operator headsets. Hardware requirements depend on how many people must collect demonstrations or intervene at the same time. Reusing training data and improving autonomy can change that relationship.
New tasks and sites could still require additional operators and equipment. Neither Noble nor NVIDIA discloses enough headset or operator data to calculate the ratio, and one developer cannot establish demand across robotics.
An XR supplier pursuing this market would need to establish who funds the demonstration stations and what they buy, repeatedly. Integration and data-management work could generate ongoing business even when the headset order is small. Forecasting sales from the number of robots deployed would miss how the training system operates.
Customers Can Outlast Their Platforms
Enterprise VR demand also offers limited reassurance about the survival of an individual supplier.
On 1 September, ENGAGE XR announced plans to enter liquidation after a major customer said it would stop taking licences. Its financial difficulties predated that decision, but the withdrawal nevertheless removed expected income from a business already dependent on its arrival.
A supplier can have recognisable customers and still depend heavily on a small number of payments. Buyers need to understand that exposure before committing content and staff training to a platform they expect to use for years.
Hardware withdrawal creates a related problem: Microsoft stopped manufacturing HoloLens in December 2024, while medical applications built for it remained commercially relevant.
Sira Medical co-founder Rick Beberman told The XR Beat in its reporting on medical AR migration:
“We’re happy to sell our HoloLens application, but health systems understandably don’t want to invest in a sunsetting product.”
MediView separately told XRB that moving its XR90 system to another headset would require new FDA submissions and a multi-month validation process. An alternative device therefore carries costs beyond its purchase price.
HoloLens is an AR example, and neither its withdrawal nor ENGAGE’s difficulties explains Counterpoint’s quarterly VR decline. However, both demonstrate why demand for an application and the continued availability of its supplier require separate scrutiny.
Charge for Keeping the Deployment Working
Counterpoint also reports 138% annual growth in AR device shipments, with media-viewing products using birdbath or flat-prism optics accounting for 60% of that market. Growth in those devices offers no automatic replacement for a fleet running immersive training or a validated medical application.
Enterprise vendors serving customers who buy hardware infrequently need a way to fund continued development and support. Paid content updates, replacement services and migration work can earn recurring business when they address costs the customer would otherwise bear internally.
Procurement teams can make the comparison concrete by asking suppliers to price support over the intended life of the deployment, including replacement equipment and the work required to move the application to another device.
Before expanding a fleet, buyers should also ask for a demonstration of content and training-record export, with an explanation of what will run elsewhere and what must be rebuilt. A successful training programme can become more expensive to move as its content library grows; perhaps the migration plan belongs in the first contract, while the customer still has a choice of suppliers.
